BenchLeader

HLE-Diamond (Scale) leaderboard

As of 16 Nov 2026, the top score on HLE-Diamond (Scale) is 72.4% by Claude Fable 5.2 (Anthropic), ahead of GPT-6 Astra at 60.6%, across 20 model configurations with a published result (Scale AI SEAL, result dated 11 Nov 2026). Scale AI and CAIS's refined 1,000-question subset of Humanity's Last Exam. Scale AI's own run; the BenchLeader Index uses Humanity's Last Exam itself. HLE-Diamond (Scale) is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 16 Nov 2026)

  • Claude Fable 5.2 holds the top HLE-Diamond (Scale) score at 72.4% (Scale AI SEAL, 11 Nov 2026).
  • GPT-6 Astra is second on HLE-Diamond (Scale) at 60.6% (Scale AI SEAL, 9 Sep 2026).
  • HLE-Diamond (Scale): Scale AI and CAIS's refined 1,000-question subset of Humanity's Last Exam. Scale AI's own run; the BenchLeader Index uses Humanity's Last Exam itself (Scale AI SEAL).
  • 20 model configurations have a published HLE-Diamond (Scale) result on BenchLeader (BenchLeader, 16 Nov 2026).
  • HLE-Diamond (Scale) does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Scale AI and CAIS's refined 1,000-question subset of Humanity's Last Exam. Scale AI's own run; the BenchLeader Index uses Humanity's Last Exam itself.

Scale AI SEAL marks 12 of these 20 results as possibly contaminated: the model was evaluated after the questions were public, so they may already be in its training data. Those scores are in the table below; this board does not feed the index.

Published by
Scale AI SEAL
Category
Reasoning
Index weight
Reference only
Models
20
Data as of
16 Nov 2026
Newest result
11 Nov 2026

Scale AI SEAL Leaderboards.

20 of 20
#
1NewClaude Fable 5.272.4%65.42026-11-11–
2GPT-6 Astra60.6%66.42026-09-09possible
3Claude Opus 5.555.0%61.02026-09-03possible
4Gemini 4 Argon53.2%64.42026-10-16–
5Claude Fable 5.151.3%66.42026-04-10possible
6Claude Opus 538.6%63.22026-09-09possible
7Claude Sonnet 5.537.5%–2026-10-15–
8Muse Wonder37.3%64.42026-10-16–
9GPT-6.1 Terra35.7%–2026-10-28–
10Gemini 3.8 Flash34.3%56.42026-03-23possible
11GPT-6 Sol33.8%–2026-04-08possible
12GPT-5.6 Sol31.2%54.32025-11-19possible
13Muse Spark Open28.0%–2026-10-16–
14Muse Spark 1.325.4%61.62026-03-10possible
15Grok 4.723.4%–2026-02-17possible
16Kimi K322.2%60.72026-04-22possible
17DeepSeek V4.1 Flash 101720.5%–2026-10-28–
18GLM 5.316.4%–2025-11-06possible
19DeepSeek V4 Pro13.4%53.02025-12-15possible
20Inkling11.2%50.42026-10-13–

AI reasoning and math benchmark leaderboard · All LLM benchmarks

Cite as: BenchLeader, “HLE-Diamond (Scale) leaderboard”, data as of 16 Nov 2026, https://www.benchleader.com/benchmarks/scale_hle_diamond.

HLE-Diamond (Scale): questions

What does HLE-Diamond (Scale) measure?
Scale AI and CAIS's refined 1,000-question subset of Humanity's Last Exam. Scale AI's own run; the BenchLeader Index uses Humanity's Last Exam itself. Scores are reported in percent of tasks solved; higher is better.
What is the top score on HLE-Diamond (Scale)?
As of 16 Nov 2026, the highest score on HLE-Diamond (Scale) is 72.4% by Claude Fable 5.2 (result dated 11 Nov 2026), ahead of GPT-6 Astra at 60.6%. Scores come from Scale AI SEAL and are re-read daily.
Which AI model leads HLE-Diamond (Scale)?
Claude Fable 5.2 leads HLE-Diamond (Scale) with 72.4% as of 16 Nov 2026 (result dated 11 Nov 2026), ahead of GPT-6 Astra at 60.6%.
How many models have HLE-Diamond (Scale) results?
20 model configurations have a HLE-Diamond (Scale) result on BenchLeader, all taken from Scale AI SEAL.
Who runs HLE-Diamond (Scale) and how often is it updated?
HLE-Diamond (Scale) is published by Scale AI SEAL. BenchLeader re-reads the published results once a day and records the date each result was published.
Does HLE-Diamond (Scale) count toward the BenchLeader Index?
No. HLE-Diamond (Scale) is shown for reference but left out of the composite index.